[Gretl-users] Re: chow test
On Wed, 31 Jan 2024, Alison Loddick wrote:
I’m hoping the econometricians can help me. I’m working with a
colleague at another university and he recommends using the Chow
test to test whether we should use an OLS model vs Fixed effects.
My question is what test does Gretl use within Fixed effects to
compare the model with OLS as the manual doesn’t say?
The manual recommends that we use the fixed effect model results
to see whether there should be a fixed effects model or OLS. I
assumed this was the Chow test. My output is below:
Test for differing group intercepts -
Null hypothesis: The groups have a common intercept
Test statistic: F(98, 1230) = 11.49
with p-value = P(F(98, 1230) > 11.49) = 4.27683e-116
This is a standard Wald F-test, supporting the conclusion that
(basically) there's no way the groups have a common intercept, or in
other words plain OLS is not an acceptable estimator. Concepts
needed: null hypothesis, p-value.
There is a Chow test on the OLS model but when I run it has
different degrees of freedom and different F-statistic.
I could spell out what the Chow test does, but here's a better idea,
along the lines of giving someone a fish versus teaching them how to
fish: how about reading the help text for the "chow" command in
gretl? You should find out it's about a different matter altogether.
Allin Cottrell